Professional Documents
Culture Documents
This chapter shows the solution method for Greens functions of 1, 2 and 3D
Laplace and wave equations. Lengthy and detailed explanations are given in order
to instruct the basic technique of the integral transform. Especially, Fourier
inversion integral for the time-harmonic Greens function is discussed in detail.
Now, we start from the simplest wave equation that has two variables: a single
space variable x and the time t,
o2 / 1 o2 /
Pdxdt 2:1:1
ox2 c2 ot2
The nonhomogeneous term represents a wave source with magnitude P, and two
Diracs delta functions, dx and dt, show the location and the impulsive nature
of the source. The Greens function is a particular solution of the differential
equation corresponding to the impulsive source. The Greens function is sought
under the quiescent condition at an initial time,
o/
/jt0 0 2:1:2
ot t0
and the convergence condition at infinity,
o/
/jx!1 0 2:1:3
ox x!1
To obtain a particular solution of the wave equation (2.1.1), we apply the
integral transforms and reduce the differential equation to an algebraic equation in
the transformed domain. Since the unknown function /x; t has two variables, we
apply the double transform: Laplace transform with respect to the time variable
t : 0 t\ 1,
Z1
f s Lf t f t expstdt 2:1:4
0
Firstly, we multiply the kernel of the Laplace transform expst to both sides
of the differential equation (2.1.1),
Z1 2
o / 1 o2 /
Pdxdt expstdt 2:1:6
ox2 c2 ot2
0
and perform the Laplace transform integral term by term. The order of integration
and differentiation are interchanged for the first term in the left hand side of the
equation. The first term in the right hand side is left in its order and then integrated
by parts. The last nonhomogeneous term, which has the delta function, is evaluated
by using the formula (1.2.3). Then, Eq. (2.1.6) is rewritten as
Z1 Z1 Z1
d2 1 o2 /
/ expstdt expstdt Pdx dt expstdt 2:1:7
dx2 c2 ot2
0 0 0
d2 /
s=c2 / Pdx 2:1:11
dx2
It is possible to obtain the exact solution for the above ordinary differential
equation by the elementary method. However, in order to demonstrate the integral
transform technique, we further apply Fourier transform to the ordinary differential
equation (2.1.11). The Fourier transform defined by Eq. (2.1.5) is applied to
Eq. (2.1.11),
Z1
d 2 / 2
s=c / Pdx expinxdx 2:1:12
dx2
1
the transform of the space-derivative in Eq. (2.1.13) is carried out with aid of the
convergence condition,
Z1 2
d /
expinxdx
dx2
1
x!1 Z1 2:1:16
d/ x!1
expinx in/ expinxx!1 n2
/ expinxdx
dx x!1
1
~
n2 /
14 2 Greens Functions for Laplace and Wave Equations
is also used for evaluating the last term. The Fourier transform of the ordinary
differential equation (2.1.12) yields a simple algebraic equation for the double
transformed unknown function / ,
s2
n2 / / P 2:1:18
c2
Then we have the exact expression for the double transformed function / ,
P
/ 2:1:19
n s=c2
2
The unknown function has just been determined explicitly in the transformed
domain. We shall carry out two inverse transforms successively. As the first
inversion, we apply the Fourier inversion integral which is defined by the second
of Eq. (2.1.5). The formal Fourier inversion is given by the integral
Z1
1 P
/ expinxdn 2:1:20
2p n s=c2
2
1
Due to the symmetric nature of the integrand, the integral is reduced to the simpler
semi-infinite integral,
Z1
P 1
/ cosnxdn 2:1:21
p n s=c2
2
0
This is a simple integral and the integration formula (Erdlyi 1954, pp. 8, 11),
Z1
1 p
cosxydx expajyj 2:1:22
x2 a 2 2a
0
ct
cP 1 1
/ L1 / L expsjxj=c 2:1:24
2 s
Fortunately, we have the Laplace inversion formula (Erdlyi 1954, pp. 241, 1),
1 0; t\a
L1 expas Ht a 2:1:25
s 1; t [ a
where H: is Heavisides unit step function. Applying this formula to Eq. (2.1.24),
the solution for the non-homogeneous 1D wave equation is obtained as
cP cP
/ Ht jxj=c Hct jxj 2:1:26
2 2
This solution shows an expanding (or out-going) 1D wave with uniform amplitude
cP=2 as shown in Fig. 2.1. Consequently, we get the Greens function for the 1D
wave equation,
o2 / 1 o2 / cP
Pdxdt ) /x; t Hct jxj 2:1:27
ox2 c2 ot2 2
o2 / 1 o2 /
Qdx expixt 2:2:1
ox2 c2 ot2
where Q is the source magnitude and x the frequency of the time-harmonic
vibration. We assume that its solution satisfies the convergence condition at
infinity, i.e.
16 2 Greens Functions for Laplace and Wave Equations
o/
/jx!1 0 2:2:2
ox x!1
As the first step of the solution method, we assume that the solution is also
time-harmonic,
d 2 /#
x=c2 /# Qdx 2:2:5
dx2
The exact solution of this ordinary differential equation can be obtained by the
elementary method. However, in order to demonstrate the integral transform
technique, we apply the Fourier transform, which is defined by Eq. (2.1.5), to the
ordinary differential equation (2.2.5),
Z1 2 #
d / 2 #
x=c / Qdx expinxdx 2:2:6
dx2
1
Z1
d 2 /# #
expinxdx n2 / 2:2:9
dx2
1
Then, Eq. (2.2.6) yields a simple algebraic equation for the Fourier transformed
amplitude function,
# x=c2 /
n2 / # Q 2:2:11
The Fourier transformed amplitude is determined completely,
# Q
/ 2:2:12
n2 x=c2
Our next task is to invert the transformed amplitude. Applying the formal
Fourier inversion integral to Eq. (2.2.12), we get
Z1
# 1 P
/ expinxdn 2:2:13
2p n2 x=c2
1
Inspecting the integrand, we see that it has two simple poles at n x=c, that is
to say, the poles are located on the integration path (real axis in the complex n-
plane). Since the integration cannot be performed through these singular points, we
have to distort the integration path around the poles. There are two ways of
deforming the path. One is through an upper semi-circle, other is through a lower
semi-circle as shown in Fig. 2.2. We have to determine which semi-circle is
suitable. Discussing the nature of the initial wave equation (2.2.1), we learn that
the wave will expand to the outer region from the source point, i. e. wave radiation
Im( )
+ / c +
? Re( )
?
/ c
Fig. 2.2 Possible deformations of the integration path around the pole
18 2 Greens Functions for Laplace and Wave Equations
from the source. Therefore, we have to choose the path so that the inversion
integral results in the radiation (out-going) wave from the source.
Still, it is somewhat complicated to explain the path selection. To aid the
understanding, two integrals with complex frequency are considered. After dis-
cussing the wave nature derived from each integral, we will determine and
understand the path distortion.
Let us introduce and add a small imaginary number e to the frequency in
Eq. (2.2.13) so that the poles are shifted from the real axis and are not on the
integration path. The frequency is considered in two ways, positive and negative
imaginary parts, x ! - ie. Employing the complex frequency, we consider the
complex integral,
Z
1 P
U expinxdn 2:2:14
2p n2 x=c2
C
where the integrand is the same as that in the Fourier inversion integral (2.2.13),
but the frequency is complex, x - ie. The integration loop C for the two cases
of complex frequency, with positive and negative imaginary parts, is discussed
separately.
(1) Small positive imaginary: x - ie
When the frequency has a small positive imaginary part, the poles are shifted
from the real axis. The integration path C is chosen so that the integrand vanishes
on the large semi-circle with infinite radius. We employ the lower closed loop C
in the case of positive x and the upper loop C in that of negative x as shown in
Fig. 2.3.
x<0
( + i ) / c
Re( )
( + i ) / c
x>0
C ()
2.2 1D Time-Harmonic Source 19
On the other hand, when x\0, we employ the upper loop C for the complex
integral U and have
Z1 " #
1 P 2pi Pn x=c
expinxdn expinx 2:2:17
2p n2 x=c2 2p n2 x=c2
nx=c
1
Unifying the two Eqs. (2.2.16) and (2.2.18), we have for the Fourier inversion
integral, where the frequency has a positive imaginary part, i.e.
Z1
1 P iP
2 2
expinxdn expixjxj=c ; x - ie
2p n x=c 2x=c
1
2:2:19
x<0
( i ) / c
Im( )
( i ) / c
x>0
C ()
Z1 " #
1 P 2pi Pn x=c
expinxdn expinx
2p n2 x=c2 2p n2 x=c2
nx=c
1
2:2:20
Rewriting the above, we have in the case of positive x,
Z1
1 P iP
2 2
expinxdn expixx=c ; x [ 0 2:2:21
2p n x=c 2x=c
1
Unifying two Eqs. (2.2.21) and (2.2.23), we have for the frequency with the
negative imaginary part,
2.2 1D Time-Harmonic Source 21
Z1
1 P iP
2 2
expinxdn expixjxj=c ; x - ie
2p n x=c 2x=c
1
2:2:24
Two expressions are obtained for the single Fourier inversion integral. They are
Eqs. (2.2.19) and (2.2.24) and are summarized in a unified expression as
(
iP
2x=c expixjxj=c; x - ie
/# iP 2:2:25
2x=c expixjxj=c; x - ie
The wave nature of the two expressions is discussed by multiplying the time
factor,
(
iP
2x=c expfixt jxj=cg ; x - ie
/ iP 2:2:26
2x=c expfixt jxj=cg ; x - ie
Inspecting the argument of the exponential function in the above Eq. (2.2.26), the
upper solution for the negative imaginary part gives an out-going (radiation) wave
from the source, but the lower solution for the positive imaginary part gives an
incoming wave, coming from infinity. Since a suitable solution must have the out-
going wave behavior, we employ the upper formula in Eqs. (2.2.25) and (2.2.26).
Thus, the frequency with negative imaginary part is the suitable assumption. Then,
the suitable integration contour for the Fourier inversion integral is that of Fig. 2.4
and the final result for the time-harmonic response is given by
o2 / 1 o2 / iQ
2
2 2 Qdx expixt ) / expfixt jxj=cg
ox c ot 2x=c
2:2:27
We have just learned that the proper selection for a complex frequency is
x - ie, i.e. negative imaginary part; and the suitable integration loop is C
in Fig. 2.4. If we do not introduce a small imaginary part and keep the integration
path and the location of the poles on the real axis, the integration path around the
pole should be deformed by a small semi-circle shown in Fig. 2.5. This defor-
mation is only valid in the case of a positive time factor, expixt. If we use a
negative time factor expixt, the integration path on the real axis must be
deformed by that shown in Fig. 2.6. Thus, the selection of the deformed path
around the pole depends on the sign of the frequency. Finally, we could answer to
the initial question about the deformation of the integration path for the Fourier
inversion integral.
22 2 Greens Functions for Laplace and Wave Equations
Im( )
Integration path
exp(+i t)
+ / c +
Re( )
/ c
Integration path
Im( )
exp(i t)
+ / c +
Re( )
/ c
Let us consider Greens function for a typical partial differential equation, the so-
called Laplace equation. The Laplace equation with a source S is the nonhomo-
geneous differential equation,
o2 / o2 /
Sdxdy 2:3:1
ox2 oy2
The product of two delta functions in the nonhomogeneous term shows the
location of the source, i.e. the source S is placed at the coordinate origin 0; 0 in
x; y-plane. The convergence condition at infinity,
o/ o/
/jpx 0 2:3:2
2 y2 !1
ox px
2 y2 !1 oy px
2 y2 !1
is also imposed.
Now, we apply the integral transforms to Eq. (2.3.1). Since the unknown
function / has two space variables, we apply the double Fourier transform defined
by
2.3 2D Static Source 23
Z1 Z1
1
/n /x expinxdx; /x /n expinxdn 2:3:3
2p
1 1
Z1 Z1
~ 1 ~
/g /y expigydy; /y /g expigydg 2:3:4
2p
1 1
With aid of the convergence condition (2.3.2), each term is transformed as follows:
0 1 1
Z1 Z 2
@ o / ~
expinxdxA expigydy n2 /
ox2
1 1
0 1 1
Z1 Z 2
@ o / ~
expinxdxA expigydy g2 / 2:3:6
oy2
1 1
0 1 1
Z1 Z
@ Sdxdy expinxdxA expigydy S
1 1
~ S for the double
Then, we have the simple algebraic equation n2 g2 /
transformed function. and its solution is given by
~
S
/ 2
2:3:7
n g2
The reader will find that the partial differential equation (2.3.1) is transformed
to a simple algebraic equation. There is thus no need of solving a differential
equation directly. The subsequent inversion process is however crucial for the
solution. The formal Fourier inversion integral with respect to the parameter g,
Z1 Z1
1
/
S
expigydg
S 1
cosgydg 2:3:8
2p 2 2 p 2
n g n g2
1 0
S expjnjjyj
/ 2:3:9
2jnj
24 2 Greens Functions for Laplace and Wave Equations
The next inversion is to evaluate the Fourier inversion integral with respect to
the parameter n,
Z1
1 S
/ expjnjjyj expinxdn 2:3:10
2p 2jnj
1
Inspecting the integrand, the singular point at n 0 lies on the real axis, i.e. on the
integration path. It is impossible to evaluate the integral in the regular sense. So,
we have to deform the integration path around the pole as in the previous section.
But it was somewhat complicated to determine the path deformation. In order to
avoid this troublesome work, we consider a simpler way of inverting the integral.
Since the trouble stems from the singular point at n 0, in order to avoid the
trouble, we differentiate Eq. (2.3.10) with respect to the space variables, x and y,
respectively,
Z1
o/ S in
expjnjjyj expinxdn 2:3:11
ox 4p jnj
1
Z1
o/ S
sgny expjnjjyj expinxdn 2:3:12
oy 4p
1
Z1
o/ S
sgny expnjyj cosnxdn 2:3:15
oy 2p
0
The two integrals in the above equations are well-known from Calculus and we
have the formulas,
Z1 Z1
x y
expny sinnxdn 2 ; expny cosnxdn 2:3:16
x y2 x2 y2
0 0
2.3 2D Static Source 25
c0 y c00 x 2:3:19
This condition is satisfied only when the two terms are pure constant and do not
include any space variable:
o2 / o2 / S
Sdxdy ) / log x2 y2 arbitrary constant
ox2 oy2 4p
2:3:21
This Greens function does not satisfy the convergence condition at infinity. This
is because we could not carry out the Fourier inversion integral of Eq. (2.3.10)
directly.
o2 / o2 / 1 o2 /
Pdxdydt 2:4:1
ox2 oy2 c2 ot2
The nonhomogeneous term states that the source with magnitude P is placed at the
coordinate origin and is impulsive in time. The quiescent condition at an initial time,
o/
/jt0 0 2:4:2
ot t0
26 2 Greens Functions for Laplace and Wave Equations
are also employed. As the wave equation (2.4.1) has two space variables x and y,
and one time variable t, a triple integral transform is applied to the differential
equation (2.4.1): the Laplace transform with respect to the time variable,
Z1
/ s /t expstdt 2:4:4
0
and the double Fourier transform with respect to the space variables,
Z1 Z1
1
/n /x expinxdx; /x /n expinxdn 2:4:5
2p
1 1
Z1 Z1
~ 1 ~
/g /y expigydy; /y /g expigydg 2:4:6
2p
1 1
Applying this triple integral transform with the quiescent and convergence con-
ditions, the original differential equation (2.4.1) is transformed to the simple
algebraic equation for the unknown function / ~
,
~ ~
g2 /
n2 / ~ P
s=c2 / 2:4:7
Then, the exact expression for /~
in the transformed domain is given by
~
P
/ 2:4:8
2
n g2 s=c2
For the inversion, three inversion integrals must be carried out successively.
The first one is the Fourier inversion integral with respect to the parameter g. This
is reduced to a semi-infinite integral, as
Z1 Z1
1
/
P
expigydg
P cosgy
q2 dg
2p 2 2 2 p
n g s=c
1 0 g2 n2 s=c2
2:4:9
The integral on the far right side is easily evaluated by applying the formula
(2.1.22). Then, the first Fourier inversion integral in Eq. (2.4.9) yields
2.4 2D Impulsive Source 27
q
P
/ q exp jyj n2 s=c2 2:4:10
2 n2 s=c2
and we apply the integration formula (Erdlyi 1954, pp. 17, 27) to get
Z1 p p
1
p exp c x2 a2 cosbxdx K0 a b2 c2 2:4:13
x 2 a2
0
where K0 : is the zeroth order modified Bessel function of the second kind. Then,
Eq. (2.4.12) takes the compact form
P s p
/ K0 x2 y2 2:4:14
2p c
The last inversion is the Laplace inversion. The Laplace inversion is symbol-
ically expressed as
P 1 h s p i
/ L K0 x2 y2 2:4:15
2p c
We have the suitable inversion formula (Erdlyi 1954, pp. 277, 8),
Ht a 0 ; t\a
L1 K0 as p p
1
; t[a 2:4:16
2
t a 2
t2 a2
Applying this formula to Eq. (2.4.15), we have the simple expression for /,
(
cP H ct r cP 0 ; ct\r
/ q p
1
; ct [ r 2:4:17
2p 2p
ct2 r 2 ct2 r 2
where H: is Heavisides unit step function and the radial distance r from the
source is defined by
p
r x2 y2 2:4:18
28 2 Greens Functions for Laplace and Wave Equations
o2 / o 2 / 1 o2 / cP Hct r
Pdxdydt ) / q 2:4:19
ox2 oy2 c2 ot2 2p
ct2 r 2
o2 / o2 / 1 o2 /
Qdxdy expixt 2:5:1
ox2 oy2 c2 ot2
where Q and x are magnitude and frequency of the source, respectively. We also
assume that its Greens function satisfies the convergence condition at infinity,
p o/ o/
/j x2 y2 !1 p p 0 2:5:2
ox x2 y2 !1 oy x2 y2 !1
As the standard technique for the time-harmonic response, the Greens function
is assumed as the product
/x; y; t /# x; y expixt 2:5:3
where /# x; y is an amplitude function to be determined. Substituting this
assumption into the nonhomogeneous wave equation (2.5.1), we have the reduced
wave equation (the so-called Helmholtz equation) for the amplitude function
/# x; y,
o2 /# o2 /#
x=c2 /# Qdxdy 2:5:4
ox2 oy2
The convergence condition (2.5.2) is also rewritten for the amplitude function as
o/# o/#
/# px 0 2:5:5
2 y2 !1
ox px
2 y2 !1 oy p
x2 y2 !1
The double Fourier transform with respect to two space variables as defined by
Eqs. (2.4.5) and (2.4.6) is applied to the nonhomogeneous Helmholtz equation
(2.5.4),
2 3
Z1 Z1 2 # 2 #
4 o / o /
x=c / Qdxdy expinxdx5 expigydy
2 #
ox2 oy2
1 1
2:5:6
2.5 2D Time-Harmonic Source 29
Equation (2.5.6) is transformed into the algebraic equation for the unknown
~
#,
amplitude function /
~ # ~
g2 /
n2 /
# ~ Q
x=c2 / #
2:5:8
Then, the explicit expression for the amplitude function in the transformed domain
is
~
# Q
/ 2:5:9
n2 g2 x=c2
As the first inversion, we apply the Fourier inversion integral with respect to the
parameter g. Its formal inversion integral is simplified as
Z1 Z1
# 1 Q Q cosgy
/ expigydg q2 dg
2p n2 g2 x=c2 p
1 0 g2 n2 x=c2
2:5:10
The far right integral is evaluated with aid of the formula (2.1.22) and yields
q
Q
#
/ q exp jyj n2 x=c2 2:5:11
2 2
2 n x=c
The second Fourier inversion integral with respect to the parameter n is given
by
Z1 q
Q 1
#
/ q exp jyj n2 x=c2 expinxdn 2:5:12
4p
1 n2 x=c2
The integrand in the above inversion integral has two branch points at n x=c
which are on the real axis in the complex n-plane, i.e. on the integration path.
q
Thus, the radical n2 x=c2 changes its sign around the branch point and we
have to discuss the path deformation around these as was done in the case of the
1D time-harmonic problem in Sect. 2.2. Since a detailed discussion for the
introduction of branch cuts is somewhat complicated, we shall follow the math-
ematical procedure described in the book of Ewing, Jardetzky and Press (Ewing
et al. 1957, pp. 4451) and explain it concisely.
30 2 Greens Functions for Laplace and Wave Equations
In order to evaluate the inversion integral of Eq. (2.5.12), we define the integral
I as
Z1 q
1 1
I q exp inx y n2 x=c2 dn 2:5:13
2p
1 n2 x=c2
and consider the complex integral having the same integrand as that in the above
equation,
Z q
1 1
U q exp inx y n2 x=c2 dn 2:5:14
2p 2 2
C n x=c
q
The radical n2 x=c2 has two branch points, and two branch cuts are
introduced in order to keep the radical single-valued. If we add a small negative
imaginary number to the frequency, as was done in Sect. 2.2, the branch points are
shifted from the real axis. Two branch cuts are introduced along the hyperbola as
shown in Fig. 2.7. When the imaginary part of the frequency vanishes, the branch
cut lies on the real and imaginary axes as shown by the red line in Fig. 2.8. Then,
the closed loop C for the complex integral is composed of four contours. One is a
straight line ABC along the real axis (green line), the second one is the line DEF
\
and GHI along the branch cut (blue line), a small circle FG around the branch
\ \
point and the last is two quarter circles, CD and AI , with infinite radius (dotted
curve). The integral along the small circle around the branch point vanishes as its
radius tends to zero, and the two integrals along the large arcs also vanish as the
Im( )
branch cut
arg ( ) = / 2
for x < 0
= ( / c) Re( )
arg ( ) = + / 2
= +( / c)
branch cut
Fig. 2.7 Branch cuts in the case of the negative imaginary part of the frequency
2.5 2D Time-Harmonic Source 31
Im( )
A
Fourier inversion line
F
B
E
C Re( )
H G
arg ( ) = / 2
arg ( ) = + / 2
For x > 0
D I
Fig. 2.8 Closed loop C for the complex integral U in Eq. (2.5.14)
radius tends to the infinity. It is enough to consider two dominant integration paths
for the complex integral. One is the line ABC along the real axis and other is the
lines DEF and IHG along the branch cut. The argument and value of the radical
along the branch cut is shown in Table 2.1
Then, the closed loop C for the complex integral is ABCDEFGHIA in Fig. 2.8
and we can apply Cauchys theorem to the complex integral U. As no singular
point is included within the loop, and the integrals around all circular paths vanish,
the line integral along the real axis, i.e. the Fourier inversion integral, is converted
to the integrals along the branch line DEF and GHI. For the branch line integrals,
we replace the integration variable with n g or ig, where g is real and
positive. Then, the line integral is converted to real valued integrals, as
Z1 q
1 1
q exp inx y n2 x=c2 dn
2p
1 n2 x=c2
Z1 q
1 1
q cos y g2 x=c2 expgxdg
p
0 g2 x=c2
2:5:15
Zx=c q
1 1
q cos y x=c2 g2 singxdg
p
0 x=c2 g2
Zx=c q
i 1
q cos y x=c2 g2 cosgxdg
p
0 x=c2 g2
32 2 Greens Functions for Laplace and Wave Equations
The Fourier inversion integral has just been converted to three real integrals.
Fortunately, we can evaluate these integral with aid of integration formulas. Two
integration formulas (Gradshteyn and Ryzhik 1980, pp. 755, 6.677, No. 4)*
Z1 p Za p
1 1
p cos b a2 x2 expcxdx p cos b a2 x2 sincxdx
2
a x 2 a 2 x2
0 0
p p
Y0 a b2 c2
2
2:5:16
and (Erdlyi 1954, pp. 28, 42)
Za p
1 p p
p cos b a2 x2 coscxdx J0 a b2 c2 2:5:17
a2 x 2 2
0
are applied to the right hand side of Eq. (2.5.15) and the Hankel function (Watson
1966, pp. 73) is also introduced. Equation (2.5.15) is simplified as
Z1 q
1 1
q exp inx y n2 x=c2 dn
2p
1 n2 x=c2
1 x p i x p 2:5:18
Y0 x2 y2 J0 x2 y2
2 c 2 c
i 2 x p
H0 x2 y2
2 c
where J0 : and Y0 :are Bessel functions of the first and second kind, respectively.
Finally, we have a simple expression for the Fourier inversion integral. That is
2.5 2D Time-Harmonic Source 33
Z1 q
1 1
q exp inx y n2 x=c2 dn
2p
1 n2 x=c2
i 2 x p
H0 x2 y2 2:5:19
2 c
It will be shown later that the above equation is valid for the case of the positive
frequency, expixt.
2:5:20
if we consider the above as the Fourier cosine transform with the transform
parameter b, its inverse cosine transform is given by
Za q
p p 1
Y0 a x2 z2 p sin z a2 b2 cosbxdb
2 a 2 b2
0
2:5:21
Z1 q
1
p exp z b2 a2 cosbxdb
b2 a2
a
p
Introducing the variable transform, u b2 a2 for the second integral in the
right hand side, the formula (2.5.16) is obtained as
Za q
p p 1
2
Y0 a x z 2 p sin z a2 b2 cosbxdb
2 a 2 b2
0
2:5:22
Z1 p
1
p expzu cos x u2 a2 du
u2 a2
0
The Fourier inversion integral has just been evaluated and is expressed by the
Hankel function of the second kind and the amplitude function of Eq. (2.5.12) is
expressed in a compact form as
iQ 2 x p
/# H0 x2 y2 2:5:23
4 c
Thus the 2D Greens function for the time-harmonic source is given by
34 2 Greens Functions for Laplace and Wave Equations
o2 / o2 / 1 o2 /
Qdxdy expixt )
ox2 ox2 c2 ot2 2:5:24
iQ 2
/ H0 xr=c expixt
4
where r is the 2D radial distance from the source, defined by Eq. (2.4.18).
Let us examine our Greens function to see if it satisfies the radiation condition
at infinity. Assuming a large distance from the source, i.e. r ! 1, we employ the
asymptotic formula for the Hankel function (Watson 1966, pp. 198),
r
2 2
Hm z expfiz mp=2 p=4g ; z ! 1 2:5:25
pz
The asymptotic form of the Greens function (2.5.24) is thus given by
r
iQ 2c
/ expfixt r=c ip=4g ; r ! 1 2:5:26
4 pxr
It is clear that this equation shows an out-going wave from the origin, since the
argument of the exponential function includes only the wave propagation character
t r=c. Furthermore, its amplitude decays with the inverse square root of the
radial distance. Thus, our selection of the closed loop is correct and the Greens
function of Eq. (2.5.24) is the suitable solution for our wave equation with a time-
harmonic source.
The static 3D Greens function for the Laplace equation is a particular solution of
the nonhomogeneous differential equation,
o 2 / o2 / o2 /
2 Sdxdydz 2:6:1
ox2 oy2 oz
where S is magnitude of the source, which is placed at the coordinate origin
0; 0; 0. The convergence condition at infinity,
p o/ o/ o/
/j x2 y2 z2 !1 p p p 0
ox x2 y2 z2 !1 oy x2 y2 z2 !1 oz x2 y2 z2 !1
2:6:2
is applied to the Greens function /.
The Greens function is obtained by the method of integral transform. For three
space variables, the triple Fourier transform defined by
2.6 3D Static Source 35
Z1 Z1
1
/n /x expinxdx; /x /n expinxdn 2:6:3
2p
1 1
Z1 Z1
~ 1 ~
/g /y expigydy; /y /g expigydg 2:6:4
2p
1 1
Z1 Z1
^ 1 ^ expifzdf 2:6:5
/f /z expifzdz; /z /f
2p
1 1
2:6:6
Applying the convergence condition (2.6.2), the above Eq. (2.6.6) is transformed
^~
to the algebraic equation for the triple transformed function /,
^
~
S
n2 g2 f2 / 2:6:7
Thus, the transformed function is determined explicitly,
^
~
S
/ 2:6:8
n g2 f 2
2
The above integral is easily evaluated with aid of the formula (2.1.22) which yields
q
~ S
p
/ exp jzj n2 g2 2:6:10
2 n2 g2
The second inversion integral with respect to the parameter g is reduced to the
semi-infinite integral
36 2 Greens Functions for Laplace and Wave Equations
Z1 q
1
/
S
p exp jzj n2 g2 expigydg
2p 2 n2 g2
1
2:6:11
Z1 q
S 1
p exp jzj n2 g2 cosgydg
2p n2 g2
0
Fortunately, we have the integration formula, (Erdlyi 1954, pp. 49, 40)
Z1
p
K0 an cosbndn p 2:6:14
2 a2 b2
0
Applying this formula to Eq. (2.6.13), the last inversion integral is exactly eval-
uated as
S
/ p 2:6:15
4p x y2 z2
2
Consequently, we have the static 3D Greens function for the Laplace equation,
o2 / o2 / o 2 / S
2 Sdxdydz ) / 2:6:16
ox2 oy2 oz 4pR
where R is the 3D radial distance from the source,
p
R x2 y2 z2 2:6:17
2.7 3D Impulsive Source 37
Greens function for the 3D wave equation is discussed. The wave equation with
an impulsive point source located at the coordinate origin is given by
o2 / o2 / o2 / 1 o2 /
2
2 2 2 2 Pdxdydzdt 2:7:1
ox oy oz c ot
where P is the magnitude of the source. The quiescent condition at an initial time,
o/
/jt0 0 2:7:2
ot t0
and the convergence condition at infinity,
p o/ o/ o/
/j x2 y2 z2 !1 p p p 0
ox x2 y2 z2 !1 oy x2 y2 z2 !1 oz x2 y2 z2 !1
2:7:3
are also imposed.
Our dynamic Greens function has four variables: three space and one time
variables. Laplace transform with respect to the time,
Z1
/ s L/t /t expstdt 2:7:4
0
and the triple Fourier transform with respect to three space variables,
Z1 Z1
1
/n /x expinxdx; /x /n expinxdn 2:7:5
2p
1 1
Z1 Z1
~ 1 ~
/g /y expigydy; /y /g expigydg 2:7:6
2p
1 1
Z1 Z1
^ 1 ^ expifzdf 2:7:7
/f /z expifzdz; /z /f
2p
1 1
are applied to the nonhomogeneous wave equation (2.7.1). With aid of the qui-
escent and convergence conditions, the wave equation is transformed to the simple
^~
algebraic equation for the multi-transformed unknown function / ,
^~
fn2 g2 f2 s=c2 g/ P 2:7:8
38 2 Greens Functions for Laplace and Wave Equations
Z1 q p
p
K0 a n2 b2 coscndn p exp b a2 c2 2:7:15
2 a2 c 2
0
is very helpful for our task. Then, applying the above formula (2.7.15) to the last
integral in Eq. (2.7.14), we have
P s p
/ p exp x 2 y 2 z2 2:7:16
4p x2 y2 z2 c
o2 / o2 / o2 / 1 o2 / P
Pdxdydzdt ) / dt R=c
ox2 oy2 oz2 c2 ot2 4pR
2:7:20
This section derives the 3D Greens function for a time-harmonic source. It is the
convolution integral of the impulsive Greens function obtained in the previous
section. The wave equation with the time-harmonic source is given by
o2 / o2 / o2 / 1 o2 /
Qdxdydz expixt 2:8:1
ox2 oy2 oz2 c2 ot2
where Q is magnitude of the source and x frequency of the time-harmonic
vibration. The convergence condition at infinity,
40 2 Greens Functions for Laplace and Wave Equations
o/ o/ o/
/jp
x2 y2 z2 !1
p p p 0
ox x2 y2 z2 !1 oy x2 y2 z2 !1 oz x2 y2 z2 !1
2:8:2
is imposed.
The standard multiple integral transform technique is available for getting the
time-harmonic Greens function. However, as shown in the case of 2D Greens
function in Sect. 2. 5, a tiresome complex integration must be carried out. In order
to avoid this difficulty, we take a very simple way. That is, the convolution integral
of the Greens function for the impulsive source.
In the previous section we get the Greens function for the impulsive source as
o2 / o2 / o2 / 1 o2 / P
2
2
2
2 2
Pdxdydz ) / dt R=c 2:8:3
ox oy oz c ot 4pR
Replacing the source magnitude P with the time-harmonic function with the fre-
quency x,
P ! Q expfixt t0 g 2:8:4
and denote the time variable in the Greens function (2.8.3) with t0 . The convo-
lution integral yields
Zt
Q
/ dt0 R=c expfixt t0 gdt0 2:8:5
4pR
0
It is very easy to evaluate the above integral, since the integrand includes Diracs
delta function and we can apply the simple integration formula (1.2.3) in Sect. 1.2,
Zb
f c ; a\c\b
f xdx cdx 2:8:6
0 ; c\a or b\c
a
Then, we can evaluate the integral in Eq. (2.8.5) and have for /
Q
/ Ht R=c expfixt R=cg 2:8:7
4pR
The step function ahead of the equation means that Eq. (2.8.7) is the transient
response to the time-harmonic source and the disturbance starts from the wave
arrival t R=c. When sufficient long time has passed and the response becomes
steady, the step function is meaningless. Then, we have the steady-state time-
harmonic response as
Q
/ expfixt R=cg 2:8:8
4pR
Table A Greens function for Laplace and Wave equations
Differential equation Source Greens function
2
1D o2 / Pdxdt /x; t cP
ox2 c12 oot/2 Source 2 Hct jxj
2.8 3D Time-Harmonic Source
Qdx expixt iQ
/x; t 2x=c expfixt jxj=cg
2 p
2D o2 / Sdxdy S
ox2 ooy/2 Source /x; y 2p logr arbitrary constant ; r x2 y2
2 2
p
o2 / Pdxdydt cP
2
ox2 oox/2 c12 oot/2 Source /x; y; t p
H ct r ; r x2 y2
2
2p ct r
2
p
Qdxdy expixt /x; y; t iQ r x2 y2
4 H0 xr=c expixt ;
2 2
p
3D o2 / Sdxdydz S
ox2 ooy/2 ooz/2 Source /x; y; z 4pR ; R
x2 y2 z 2
2 2 2
p
o2 / Pdxdydzdt P
ox2 ooy/2 ooz/2 c12 oot/2 Source /x; y; z; t 4pR dt R=c ; R x2 y2 z2
Q
p
Qdxdydz expixt /x; y; z; t 4pR expfixt R=cg ; R x2 y2 z2
41
42 2 Greens Functions for Laplace and Wave Equations
The 3D Greens function for the wave equation with the time-harmonic source is
given by
o2 / o2 / o2 / 1 o2 /
2
2 2 2 2 Qdxdydz expixt
ox oy oz c ot 2:8:9
Q
) / expfixt R=cg
4pR
Exercises
(2.1) Using the impulsive response (2.1.27) for 1D wave equation, derive the
time-harmonic Greens function through the convolution integral and com-
pare it with the time-harmonic Greens function (2.2.27).
(2.2) Using the impulsive Greens function (2.4.19) for 2D wave equation, derive
the time-harmonic Greens function (2.5.23) through the convolution
integral.
References
Erdlyi A (ed) (1954) Tables of integral transforms, vol I. McGraw-Hill, New York
Ewing WM, Jardetzky WS, Press F (1957) Elastic waves in layered media. McGraw-Hill, New
York
Gradshteyn IS, Ryzhik IM (1980) In: Jefferey A (ed) Table of integrals, series, and products, 5th
edn. Chap.6, Academic Press, San Diego
Watson GN (1966) A treatise on the theory of bessel functions. Cambridge University Press,
Cambridge
http://www.springer.com/978-3-319-00878-3